Comparison of GARCH and Data Science Models in Financial Times Series Forecasting: An analysis for the financial market volatility in the banking and automobile industry.

Contents

  1. Libraries
  2. Data preparation
  3. EDA
  4. Building Models GARCH models

Data Science models

  1. Models comparison

Libraries

Data preparation

Import Data

We use data from four international stock markets, including two different industries (Banking and Automotive) retrieved from YahooFinance from the period 2000-01-01 to 2020-12-31.

Daily returns

Realized Volatility

Data per company

EDA

Plot Daily returns / historical price (per company)

DBK

BAC

We can apply the same analysis to the BAC case. However, in the United States, we can judge that the effects of the 2003 financial crisis were minimal. Indeed, we can see that prices continued to rise during this period. The same is true for the 2011-2012 period. The two major events where we notice a strong decrease in prices, as well as a high level of volatility (BAC: Daily returns) are the global crisis of 2007 and the COVID-19 health crisis.

BMW

FORD

Ford's prices dropped drastically from 2002-2003 until they reached their minimum in 2009 due to the global crisis in 2007. After this period, the prices started to increase and then decreased in 2011, which means that the company was affected by the 2011-2012 crisis that occurred in Germany. Thereafter, prices continued to decline. this analysis is confirmed by the high volatility of returns during these periods.

Plotting Daily returns (Per Industry)

DBK vs BAC

We can see that DBK was affected by the 2003 crisis, unlike LAC. However, the financial crisis of 2008 had a much greater impact on BAC than on DBK.

BMW vs FORD

We can note that the level of volatility of Ford is higher than that of BMW. We can explain this by the fact that Ford was more affected by the crises that took place during the period studied.

Building Models

GARCH models

ARCH

Deutsche Bank AG
Bank of America Corp
BMW
Ford

GARCH model

Deutsche Bank AG
Bank of America Corp
BMW
Ford

GARCH-t model

Deutsche Bank AG
Bank of America Corp
BMW
Ford

EGARCH model

Deutsche Bank AG
Bank of America Corp
BMW
Ford

GJR-GARCH model

Deutsche Bank AG
Bank of America Corp
BMW
Ford

Data Science models

LSTM

Deutsche Bank AG
Bank of America Corp
BMW
Ford

Random Forest

Deutsche Bank AG
Bank of America Corp
BMW
Ford

Models comparison

For each company, we designed and applied GARCH, LTSM, and Random Forest models to demonstrate their importance in forecasting the volatility of these markets. Based on the forecasting results obtained, we can conclude that GARCH models performed much better than Data Science models. Indeed, the best model performance was from GJR GARCH by giving the lowest RMSE and LSTM had the worst performance in both industries.